A Macroscopic Traffic Simulation Model to Mingle Manually Operated and Self-driving Cars

As the rapid development of AI, computer vision and automatic control technologies, self-driving cars have been well designed and developed. Since self-driving cars should coexist efficiently with human-driving cars, how to make practical strategies for them is increasingly significant. This paper optimizes cellular automaton to do simulation and quantizes the human factors as realistic and comprehensive as possible based on spectral clustering which is very suitable for large-scale simulation and crowd management for future smart cities. Compared with traditional analysis which record trajectories of cars, the new model employs unsupervised learning to augment average speed and reduce collision time by realizing algorithm optimization to reduce complexity and computational cost. This paper not only demonstrates the progress and results of traffic simulation, but also illustrates the concrete strategies for both self-driving cars and human drivers.